When an ecommerce firm uses artificial intelligence to accurately predict what consumers will buy with 90% accuracy, marketers, executives and entrepreneurs need to pay attention.
The Economist reported in April that German ecommerce firm Otto did just that. The company repurposed a deep learning algorithm previously used by the CERN particle physics laboratory to gaze into its customers’ commercial futures. The algorithm inhaled data on nearly three billion past transactions and used 200 different variables to predict what customers would order—before they bought it.
“The AI system has proved so reliable—it predicts with 90% accuracy what will be sold within 30 days—that Otto allows it automatically to purchase around 200,000 items a month from third-party brands with no human intervention,” The Economist reports.
This application of AI is as practical as it gets. Otto’s analysis showed that customers both were less likely to return items if they arrived within two days but also didn’t like shipments arriving in multiple boxes.
This created a conundrum for the company: to ship their products (which come from third-parties) faster, they needed to hustle them out the door when they were available. But that meant multiple packages, which irked consumers.
“The typical solution would be slightly better forecasting by humans of what customers are going to buy so that a few goods could be ordered ahead of time,” reports The Economist. Otto didn’t go with the typical solution. The company built an artificial intelligence system to handle the problem “using the technology of Blue Yonder, a startup in which it holds a stake.” The technology was previously used in particle physics experiments.
The result? The surplus stock that Otto must have on-hand has declined 20%. Product returns, for which Otto eats the shipping costs, have decreased by two million items per year.
Businesses have access to more data than at any time in past human history. Artificial intelligence offers marketers, executives and entrepreneurs the ability to surface insights from these massive datasets that humans just aren’t able to find on their own. Using those insights, machines and humans are working together to predict, personalize and produce like never before.
Otto did it, and now the company has a competitive advantage that firms not using AI just can’t match.
Ecommerce is ripe for AI disruption. If you work in it—or for clients that use it—artificial intelligence needs to be on your radar.
The Otto story highlights the vast potential for firms who understand and apply AI solutions to their businesses. It also points to important opportunities and challenges for marketers getting started with AI, machine learning, deep learning and other related technologies.
1. AI is here to make your existing work better, not take your job.
“Otto did not fire anyone as the result of its new algorithmic approach: it hired more, instead,” notes The Economist. Otto’s algorithms did something humans couldn’t, so that the bottom line improved and people were able to focus on more important work than mitigating returns. The same goes for marketing.
There’s no doubt some AI applications in marketing and other industries will disrupt traditional roles. But AI’s bigger potential lies in automating lower-level tasks, freeing marketers up to perform higher-value tasks, and augmenting the work marketers do to make them more productive and performance-driven.
Consider how much time your marketing team spends drafting social media updates, curating content, building email workflows, predicting opens, clicks and conversions, writing performance reports, and more. Now imagine if machines could perform some or all of these activities, freeing up marketers to enhance rather than create.
2. You don’t have to be Amazon (or Otto) to get started.
We love Amazon’s commitment to AI and machine learning, but you don’t have to be the world’s biggest ecommerce company to benefit from this technology. Notes The Economist, “Otto’s experience also underlines that ordinary companies can use AI, not just giants.”
Of course, Otto had a stake in an AI company, which makes adoption a lot easier. But marketers don’t need an equity stake in a machine learning outfit to get started. Plenty of AI solutions are available to marketers for free trials and/or modest monthly fees. Subscribing to the Marketing Artificial Intelligence Institute is the best way to learn about them: we regularly publish exclusive interviews with top AI marketing providers. It’s easier to bake AI into your operations than you think.
3. Focus on the outcomes you wish to achieve.
It’s easy to get bogged down in AI terminology and minutiae. (Skip the noise: we cut through the clutter in this post.) Marketers should focus instead on the outcomes they want to achieve, then speak with AI providers to determine if those outcomes could be best achieved with machine intelligence.
Otto began with a concrete challenge: how to reduce returns. Conventional data analysis was used to attack the problem, determining that returns went down if packages arrived in two days. But it also showed that customers didn’t want more packages. Otto found that artificial intelligence was perfectly suited to this challenge. It could predict far better than humans, and at scale, what customers would buy, before they bought it.
The old saying goes that to a man with a hammer, every problem looks like a nail. Don’t fall into that trap: AI can create value in many areas of your business. It is not, however, appropriate for every single scenario. Start with what you want to achieve, and proceed to investigate solutions from there.